System-on-a-chip comprising at least one secure IOMMU

The IOMMU unit with a reservation bit and dynamic entry allocation mechanism in the SoC effectively prevents covert-channel attacks while maintaining performance, addressing vulnerabilities in existing SoC designs.

FR3131403B1Active Publication Date: 2024-11-29THALES SA +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
FR2021014333
Authority / Receiving Office
FR · FR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-12-23
Publication Date
2024-11-29
Estimated Expiration
2041-12-23

AI Technical Summary

Technical Problem

Existing systems-on-chip (SoCs) with IOMMU units are vulnerable to covert-channel attacks, where peripherals exploit the TLB buffer to communicate covertly, and existing solutions either fail to protect against these attacks or significantly degrade performance.

Method used

Implement a system-on-chip with an IOMMU unit that includes a translation buffer with a reservation bit for each entry, managed by a logic circuit to dynamically allocate and reallocate entries sporadically, preventing unauthorized access collisions and maintaining performance.

Benefits of technology

The solution effectively limits covert-channel attacks while maintaining performance comparable to conventional IOMMU units, reducing the attack surface and ensuring secure memory access without significant throughput degradation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000016_0000
    Figure 00000016_0000
  • Figure 00000016_0001
    Figure 00000016_0001
  • Figure 00000017_0000
    Figure 00000017_0000
Patent Text Reader

Abstract

System on chip comprising at least one unit (IOMMU) for managing access to the processor's shared memory by peripherals to the system comprising: a translation buffer (TLB) equipped, for each entry corresponding to a translation of a virtual page into a physical page enabling the translation of transactions from peripherals (PER) to the system, with an additional reservation bit (Res) representing a reservation state or not of a translation entry of the buffer (TLB) for a peripheral;and a logic circuit (Logic) associated with the translation buffer (TLB), configured to manage the reservation bit (Res) so as to avoid overwriting a translation input of the buffer (TLB) corresponding to a translation when the reservation bit (Res) of the line has a value corresponding to a reserved state, and configured to sporadically, at bounded inter-arrival distances, reset all reservation bits of the buffer (TLB) inputs to the unreserved state. Figure for the abbreviation: Fig.2;
Need to check novelty before this filing date? Find Prior Art

Description

Title of the invention: System on chip comprising at least one secure IOMMU

[0001] The invention relates to a system on chip, or SoC for acronym for "System on Chip" in English, comprising at least one unit for managing access to the shared memory of the processor by peripherals to the system.

[0002] The invention applies to processors provided with interfaces for master peripherals (i.e. peripherals capable of making requests to the processor's memory) on the interconnection device of the elements of the system on chip, such as a communication bus, a network on chip or NoC for "Network on Chip" in English, or an interconnection network for "Interconnect" in English. The present invention ensures that the memory protection mechanism typically deployed on processors with master peripherals is not exploited as a hidden communication channel between the peripherals.

[0003] As illustrated in [Fig.l], modern systems on chip equipped with processors offer the ability to extend the functionality of the system with master peripherals PER which use the same type of communication interface or interconnection device INT as the processor core(s) to interact with the system, and in particular with the main memory MM. To prevent the PER peripherals from freely accessing the entire memory space of the system, a processor being equipped with at least one core C and possibly, or even often, at least one memory management unit MMU, places these peripherals behind a memory protection mechanism.

[0004] The component implementing this mechanism is usually called a unit for managing access to the processor's shared memory by peripherals or IOMMU for the acronym "Input-Output memory management unit" in English, but can sometimes also have the name SMMU for the acronym "System Memory Management Unit" in English, or VT for the acronym "Virtualization Technology" in English.

[0005] The IOMMU unit for managing access to the shared memory of the processor by PER peripherals to the system is the counterpart, for the peripherals, of the MMU memory management unit for the core(s) (C) of the processor.

[0006] The sharing of a main memory MM is described in an input / output page table IOPT for the PER devices and in the memory management unit page table MMUPT for the processes running on the core(s) C. These IOPT and MMUPT tables are stored in main memory MM.

[0007] The I0MMU unit for managing access to the shared memory of the processor by PER peripherals comprises a translation buffer TLB, for acronym for "Translation Lookaside Buffer" in English, each entry or line of which corresponds to a translation of a virtual address from a peripheral into a physical address intended for the system. The I0MMU unit for managing access to the shared memory of the processor also comprises a logic circuit Logic associated with the translation buffer TLB. The I0MMU unit may also comprise a current pointer module PTW, for acronym for "Page Table Walker" in English, for autonomous updating (i.e., without the intervention of the processor) of the TLB buffer.

[0008] The operating system or OS, for acronym for "Operating System" in English, configures the I0MMU unit to define the memory areas of the system that the different peripherals attached to the I0MMU unit can access. As for the MMU memory management unit, translations can be made between the addresses or SPA for "System Physical Address" in English, and the addresses of the peripherals or IOVA, for acronym for "Input / Output Virtual Address" in English, which are called virtual addresses. Other services can be provided by the I0MMU unit, but do not need to be described for the present invention.

[0009] To configure access to the shared memory provided by the I0MMU unit, the operating system uses the input / output page table IOPT in which it defines the access rights and translations for the different peripherals. This table is kept in the memory (typically DRAM) of the system. To speed up access to this table, the I0MMU unit implements a buffer memory, TLB, in which the most frequently used entries are kept.

[0010] With the I0MMU unit, the operating system can limit the memory areas of the system that peripherals can access, share a virtual address space between a process and a peripheral, and guarantee the integrity and confidentiality of the data of the different applications and peripherals of the system. This also makes it possible to ensure that two peripherals for which the operating system has defined disjoint memory areas cannot communicate with each other.

[0011] Unfortunately, peripherals can exploit the TLB buffer of the I0MMU unit to know the data that another peripheral is manipulating (side-channel attack) or to communicate data to each other (covert-channel attack). For this, approaches are also used to set up cryptographic attacks using a secondary channel that uses the synchronization of the translation buffer or "TLBleed" in English, or a hardware branch prediction vulnerability of type "Spectre" or "Meltdown" in English can be exploited.

[0012] The papers "Secure TLBs", by Shuwen Deng, Wenjie Xiong, and Jakub Szefer, Proceedings of the 46th International Symposium on Computer Architecture (ISCA), 2019, and US20200151111 Al, illustrate such state-of-the-art systems-on-chip.

[0013] As explained previously, the I0MMU unit does not protect against side-channel or covert-channel attacks by malicious devices. Implementations exist that aim to limit this type of attack, but do not directly address it, as they were proposed for reasons other than security, such as performance reasons. For example, multiple I0MMU units can be deployed in a system-on-chip, such as an I0MMU A and an I0MMU B, which prevents a device connected to I0MMU A from exploiting weaknesses in I0MMU A to transfer non-shared data (covert-channel attack) to a device connected to I0MMU B because I0MMU units are not shared. Applying this solution to the extreme by placing an I0MMU unit in front of each device would have several problems: - this is not applicable for external devices (those exposing a connector on the system-on-chip such as PCI / PCI-Express type devices) because multiple devices can be connected to the same connector. - such deployment considerably increases the surface area required on the system on chip, which increases its cost and consumption.

[0014] A solution like the one proposed for MMU memory management units, as in the documents cited above, could be implemented on the IOMMU but, in the case of adding randomness in the processing time of a memory request (see article "Secure TLBs", Proceedings of the 46th International Symposium on Computer Architecture (ISCA), 2019), no use case allows to obtain performances equivalent to an unprotected IOMMU unit. Indeed, the addition of missing translations present in the TLB buffer degrades the performances in terms of latency and throughput. This type of defense degrades the I / O performances significantly even when a single device performs memory transactions, without significantly reducing the throughput of the covert channel. Indeed, an attacker can use an error-correcting code for the potentially faulty transmitted information bits.Additionally, TLB or I0MMU partitioning techniques also fail to achieve native performance. When few devices use the unit. IOMMU, the hardware is underutilized, and the situation eventually becomes similar to the use case where one would attach one IOMMU unit per device.

[0015] State-of-the-art solutions at the MMU level are not effective in the case of a covert channel attack because an attacker can bypass them. In addition, they are not optimized for the memory traffic generated by a device.

[0016] An aim of the invention is to overcome the problems mentioned above, and in particular to limit this type of attack on the IOMMU management unit for access to the shared memory of the processor by peripherals.

[0017] According to one aspect of the invention, there is provided a system on chip comprising: - a processor with at least one core; - a shared main memory with an input / output page table; - a device for interconnecting the elements of the system; and - at least one unit for managing access to the shared memory of the processor by peripherals to the system configured to address requests to the processor comprising: - a translation buffer memory provided, for each entry corresponding to a translation of a virtual page into a physical page allowing the translation of transactions from the peripherals to the system, with an additional reservation bit representing a reservation status or not of a translation entry of the buffer memory for a peripheral; and - a logic circuit associated with the translation buffer, configured to manage the reservation bit so as to avoid replacing a translation entry of the buffer corresponding to a translation when the reservation bit of the line has a value corresponding to a reserved state, and configured to, sporadically at a bounded interarrival distance, reset all of the reservation bits of the entries of the buffer to the non-reserved state.

[0018] A sporadic event with a bounded inter-arrival distance is understood to mean a repetitive event for which a minimum limit and a maximum limit are defined for the duration separating two occurrences of the event.

[0019] According to one embodiment, the logic circuit is configured, when a translation requested by a memory request from a peripheral is present in the buffer memory, to translate the address and transfer it to the shared main memory, and test whether the reservation bit of the entry corresponds to a non-reservation, in which case the logic circuit reserves the entry for said peripheral.

[0020] In one embodiment, the logic circuit is configured, when a translation requested by a memory request from a peripheral is absent in the memory buffer, to look it up in the I / O page table using a current page pointer module of the tables, and if the translation in the I / O page table is not obtained, to raise an interrupt to the corresponding core.

[0021] According to one embodiment, the logic circuit is configured, in case of obtaining the translation in the input / output page table, to search for the presence of an invalid entry in the buffer memory, and in case of presence of an invalid entry in the buffer memory, to allocate the invalid entry of the buffer memory to said translation and write it there.

[0022] When the software controlling the unit, typically the OS, revokes access to certain memory areas for a peripheral, it removes the translation of the address in the memory I / O page table. It also sends a translation invalidation command to the peripheral access management unit. When the peripheral access management unit invalidates an entry, it invalidates the TLB buffer entry, making any further access to the data impossible.

[0023] In one embodiment, the logic circuit is configured, in case of obtaining the translation in the input / output page table, to search for the presence of an invalid entry in the buffer memory, and in case of absence of an invalid entry in the buffer memory, to search for the presence of an unreserved entry whose reservation bit corresponds to a non-reservation in the buffer memory, and in case of presence of an unreserved entry in the buffer memory, to allocate the unreserved entry of the buffer memory to said translation and write it there.

[0024] According to one embodiment, the logic circuit is configured, in the event of obtaining the translation in the input / output page table, to search for the presence of an invalid entry in the buffer memory, and in the event of the absence of an invalid entry in the buffer memory, to search for the presence of an unreserved entry whose reservation bit corresponds to a non-reservation in the buffer memory, and in the event of the absence of an unreserved entry in the buffer memory, to search for the presence of a reserved entry whose reservation bit corresponds to a reservation in the buffer memory whose transaction is being processed, to allocate the reserved entry whose transaction is being processed to said translation and write it there.

[0025] According to one embodiment, the unit for managing peripheral access to the shared memory of the system on chip is configured to be reserved for peripherals that are not predefined as trusted.

[0026] In one embodiment, the processor is also provided with at least one memory management unit, and a memory management unit page table.

[0027] The invention will be better understood by studying some embodiments described in as non-limiting examples and illustrated by the attached drawings in which the figures:

[0028] [Fig-1] schematically illustrates a system on chip comprising a processor and at least one unit for managing access to the shared memory of the processor by peripherals, according to the state of the art;

[0029] [Fig.2] schematically illustrates a system on chip comprising a processor and at least one unit for managing access to the shared memory of the processor by peripherals, according to one aspect of the invention;

[0030] [Fig.3] schematically illustrates the inputs of a buffer memory of the system on chip of [Fig.2], according to one aspect of the invention;

[0031] [Fig.4] schematically illustrates the operation of the system on chip of [Fig.2], according to one aspect of the invention;

[0032] [Fig.5] schematically illustrates the positive output of test 3 of [Fig.4], according to one aspect of the invention;

[0033] [Fig.6] schematically illustrates the negative output of test 3 of [Fig.4], according to one aspect of the invention;

[0034] [Fig.7] schematically illustrates the positive output of test 7 of [Fig.4], according to one aspect of the invention;

[0035] [Fig.8] schematically illustrates the positive output of test 9 of [Fig.4], according to one aspect of the invention;

[0036] [Fig.9] schematically illustrates the positive output of test 11 of [Fig.4], according to one aspect of the invention;

[0037] [Fig. 10] schematically illustrates the negative output of test 11 of [Fig.4], according to one aspect of the invention;

[0038] [Fig. 11] schematically illustrates a system on chip comprising a processor and at least one unit for managing access to the shared memory of the processor by peripherals, according to one aspect of the invention, and at least one state-of-the-art unit for managing access to the shared memory of the processor by peripherals;

[0039] [Fig. 12] schematically illustrates a system on chip comprising a processor and at least one unit for managing access to the shared memory of the processor by peripherals, coupled to a peripheral bus controller, according to one aspect of the invention.

[0040] Throughout the figures, elements having identical references are similar.

[0041] [Fig.2] schematically represents, according to one aspect of the invention, a system on chip comprising: - a processor equipped with at least one C core and at least one MMU memory management unit; - a shared main memory MM, for example provided with an input / output page table IOPT and a memory management unit page table MMUPT; - an INT interconnection device for the system elements; and - at least one IOMMU unit for managing PER device access to the shared memory of the system-on-chip, configured to address requests to the processor, comprising: - a TLB translation buffer provided, for each entry corresponding to a translation of a virtual page into a physical page enabling transactions from the PER peripherals to be translated to the system, with an additional Res reservation bit representing a reservation status or not of a translation entry of the TLB buffer for a peripheral; and - a logic circuit Logic associated with the translation TLB buffer, configured to manage the reservation bit Res so as to avoid replacing a translation entry of the TLB buffer corresponding to a translation when the reservation bit Res of the line has a value corresponding to a reserved state, and configured to, sporadically at a bounded inter-arrival distance, reset all the reservation bits of the entries of the TLB buffer to the non-reserved state.

[0042] The invention consists of a policy for replacing TLB buffer entries for an IOMMU unit. This policy dynamically allocates TLB buffer entries to a device for a time interval, then releases them when this time interval has elapsed.

[0043] The main advantage of the present invention is to reduce or prevent the exploitation of the covert channel that two peripherals can set up while maintaining performances very close to the nominal performances of a conventional IOMMU unit.

[0044] [Fig. 3] schematically represents the inputs of a buffer memory of the system on chip of [Fig. 2], according to one aspect of the invention.

[0045] Compared to a state-of-the-art TLB buffer memory, the TLB buffer memory of the invention is such that one of its entries comprises, in addition to the other fields, a field corresponding to a reservation bit Res.

[0046] Thus, an entry in the TLB buffer memory includes the following fields: - devicelD: identifier used by the IOMMU unit to identify the device(s) to which the translation applies; - Virtual address: virtual address of the page base used by the devices; - Physical address: physical address of the page base used for routing device requests in memory; - R: bit indicating whether or not the page (virtual or physical) can be read by the corresponding device; - W: bit indicating whether or not the page (virtual or physical) can be written to by the corresponding device; - Res: bit indicating whether or not the entry is reserved for the corresponding peripheral (devicelD).

[0047] The base of the page is understood to mean the address marking the start of a memory area of ​​predefined size.

[0048] The operating principle of the protection against covert channel attacks is to limit the possibility between two devices of creating access collisions at the level of the TLB buffer memory of the I0MMU unit.

[0049] To do this, each entry in the TLB buffer is assigned to a device for a period. A TLB buffer entry associated with a device can only have its contents replaced by other translations belonging to that same device, thereby limiting the possibility of access collisions at the TLB buffer level.

[0050] The added Res reservation bit is not accessible outside the I0MMU unit (for example by a peripheral or by the operating system running on a core), and indicates the reserved or unreserved state of a TLB buffer entry.

[0051] [Fig.4] schematically represents the operation of the system on chip of the [Fig.2], according to one aspect of the invention, or in other words the method of translating a memory transaction 1 implemented by the logic circuit Logic.

[0052] When a memory transaction 1 is received by the I0MMU unit, as illustrated in [Fig.5], the TLB buffer is scanned to find the translation of the virtual address to a physical address. The example uses a one-entry TLB buffer and in the figures, the first TLB tables before translation and the second TLB tables after translation.

[0053] If a translation is found in the TLB buffer (positive output of test 2) and if its input is reserved (positive output of test 3), the virtual address is translated into a physical address and is transferred to the main memory, as shown in [Fig.5],

[0054] The second table and the third table in [Fig.5] are identical because the translation is found in the TLB buffer, so there is no need to replace the TLB entry.

[0055] If the input that enabled the translation of the transaction is not reserved (negative output of test 3), i.e. whose reservation bit Res indicates a non-reservation, then the I0MMU unit reserves it for the corresponding peripheral, as shown in [Fig.6], by setting the reservation bit Res to reserved.

[0056] When the translation is not in the TLB buffer (negative output of test 2), the I0MMU unit scans the IOPT page table to find the translation (test 5), using a current pointer PTW module, which can be implemented in hardware or software.

[0057] If no translation is found in the IOPT page table (negative output of test 5), then the OMMU unit raises an interrupt 6 (page fault) towards core C, as illustrated in [Fig.4].

[0058] If the I0MMU unit finds the translation in the IOPT page table (positive output of test 5), it evaluates several successive conditions to choose the entry to replace in the TLB buffer, as follows.

[0059] The entry is then chosen in the following order of priority: - an invalid input (positive output of test 7), which is allocated 8, as illustrated in [Fig.7]. - a reallocable input (positive output of test 9 following negative output of test 7), which is reallocated 10, until the end of the period, as illustrated in [Fig.8]. - an entry already allocated (positive output of test 11 following negative output of test 9), as illustrated in [Fig.9] (replacing an entry already belonging to the device whose transaction is being processed). No TLB entry is available for reallocation so there is no reallocation to be done.

[0060] In the three cases 8, 10 and positive output of test 11, a backup 12 of the input / output page table IOPT is carried out, and a reservation of the entry for the peripheral carrying out the transaction.

[0061] If none of the conditions are met, (negative output of test 11) the transaction is translated and transmitted to the main memory MM without saving the translation in the TLB buffer, as illustrated in [Fig. 10].

[0062] Finally, the continuation of memory transaction 13 is carried out.

[0063] This entry allocation mechanism is complemented by a mechanism for reallocation (or re-staking) of the entries of the TLB buffer memory. This mechanism is sporadic with bounded inter-arrival distance, and consists of reinitializing all the Res bits or fields of the TLB buffer memory, by resetting them to zero (in the case where 0 means unreserved or 1 means reserved), thereby removing the reservations of all the entries of the TLB buffer memory.

[0064] However, entries whose reservation has been removed are not invalidated, and in doing so a device can reuse an entry that has been unreserved for it without suffering a TLB miss. in English. Furthermore, it is impossible for a device belonging to another device to differentiate between the retrieval of an invalid entry and the retrieval of an entry previously allocated to another device. Entries are re-allocated sporadically to prevent devices from being able to synchronize on the TLB buffer entry reallocation period to re-implement the attack with a lower throughput (the exchange frequency would become the reallocation frequency instead of being the frequency of requests processed by the I0MMU unit).

[0065] The present invention makes it possible to very strongly limit access collisions between two peripherals at the level of the TLB buffer of 1T0MMU as the peripherals are no longer capable of ejecting the translations of other peripherals out of the TLB buffer outside of the entry reallocation periods. As a result, the throughput of the hidden channel is limited, making the hidden channel unusable.

[0066] The present invention also makes it possible to maintain performances close to the native performances of a conventional I0MMU unit for conventional peripheral accesses. Also, the peripherals can be classified into two categories: - network peripherals, receiving data coming from outside the system, which are transmitted in a burst (or relatively short sequence of data but sent at the maximum speed during the window granted to the transmitter) in memory writes. The packets leaving the system are recovered by a burst of reads then transmitted outside the system. The transactions have a very high locality because they are always intended for the buffer or reception buffer or to the buffer or transmission buffer of the kernel or "kernel" in English. - calculation accelerators which load the calculation operands from the main memory in a read burst, the data is then processed by confining the execution to their internal memory of the peripheral, finally the output data is transmitted to the main memory by a write burst.

[0067] Both types of devices make memory accesses that generally have a high locality, i.e. the virtual addresses of the memory accesses are close to each other. They differ in the regularity of their accesses: the accesses of a network device are quite sporadic while the accesses of an accelerator are periodic and regular. Therefore, it is very unlikely that a device can use all the entries of a TLB buffer in a short time interval. It is very unlikely that a device will reserve an entire TLB buffer because the data that the devices manipulate often has a high locality. Therefore, several devices can operate in parallel without interference due to the system allocation of TLB buffer entries. Only malicious accesses will be impacted by this mechanism.

[0068] Also, the present invention aiming at replacing the entries of a TLB buffer memory makes it possible to: - reduce the attack surface related to covert channel attacks covert channel attacks using the translation time of a memory transaction by the IOMMU to exchange information. - to have nominal performance when peripheral memory accesses do not reserve an entire TLB buffer. - not have starvation, i.e. all device requests will be served. If a device cannot get entries in the TLB buffer, its requests are translated without being cached. We can therefore observe a performance degradation without observing starvation.

[0069] At a minimum, a system on chip with master but not trusted peripherals (internal or external) must contain an I0MMU unit according to the invention, resilient to side channel or covert channel attacks. All master peripherals (trusted or untrusted) must be connected to the rest of the system through this I0MMU.

[0070] If the area and cost constraints are not strongly limited, the system on chip can contain two I0MMU units, one resilient to attacks according to the invention, and one traditional (i.e. non-resilient), as illustrated in [Fig. 11]. On the traditional IOMMU, the trusted internal master peripherals can be connected. All other (non-master) peripherals must be connected to the resilient I0MMU unit according to the invention.

[0071] All external devices, trusted or not, must be connected to the resilient I0MMU unit. The reason being that any external device (e.g. device connected to a PCI / PCLexpress socket) can be easily replaced by another malicious device.

[0072] This implementation with a traditional I0MMU and a resilient I0MMU may allow internal devices trusted on the traditional I0MMU to avoid the slight performance degradation that the resilient I0MMU may cause.

[0073] This embodiment combining the traditional I0MMU unit and the resilient I0MMU unit according to the invention can be extended according to the surface and cost constraints of the system on chip for performance reasons, for example, with two traditional I0MMU units and one resilient I0MMU unit according to the invention, or with one traditional I0MMU unit and three resilient I0MMU units. Nevertheless, On traditional IOMMUs, only trusted internal devices should be connected. Preferably, the translation TLB buffers are multiplied in a large IOMMU, rather than multiplying the IOMMUs.

[0074] In certain cases, still for performance reasons, the resilient IOMMU unit according to the invention can be strongly coupled to a peripheral bus controller, such as for example to the root controller of a PCI / PCI-Express bus, as illustrated in [Fig. 12].

[0075] The invention is intended to be used in IOMMU units of systems on chip for which master peripherals (e.g. encryption accelerator, network interface, GPUs, TPUs, neural network, etc.) are used, and particularly when the peripherals are developed by third parties and these manipulate sensitive data.

[0076] The invention applies to systems with external peripherals (e.g. network card, graphics card, etc.), but also to systems with internal peripherals (i.e. peripherals integrated into the system-on-chip). For example, when a RISC-V based system-on-chip is developed using peripherals implemented by a third party. These peripherals can be difficult to verify because their IP blocks (for acronym for "Intellectual Property" in English) are closed (no access to its design) and / or complex IP block. Placing these peripherals behind an IOMMU unit integrating the present invention ensures that they cannot carry out side channel or covert channel attacks. This also avoids more expensive solutions such as placing a different IOMMU unit in front of each peripheral.

Claims

Claims

1. System on chip comprising: - a processor provided with at least one core (C); - a shared main memory (MM) provided with an input / output page table (IOPT); - an interconnection device (INT) for the elements of the system; and - at least one unit (IOMMU) for managing access of peripherals (PER) to the shared memory of the system on chip, configured to address requests to the processor, comprising: - a translation buffer memory (TLB) provided, for each entry corresponding to a translation of a virtual page into a physical page making it possible to translate transactions from the peripherals (PER) to the system, with an additional reservation bit (Res) representative of a reservation state or not of a translation entry of the buffer memory (TLB) for a peripheral;and - a logic circuit (Logic) associated with the translation buffer (TLB), configured to manage the reservation bit (Res) so as to avoid replacing a translation entry of the buffer (TLB) corresponding to a translation when the reservation bit (Res) of the line has a value corresponding to a reserved state, and configured to, sporadically at a bounded interarrival distance, reset all of the reservation bits of the entries of the buffer (TLB) to the non-reserved state.;

2. System according to claim 1, in which the logic circuit (Logic) is configured, when a translation requested by a memory request (1) of a peripheral is present (2) in the buffer memory (TLB), to translate the address and transfer it to the shared main memory (MM), and test (3) if the reservation bit (Res) of the entry corresponds to a non-reservation, in which case the logic circuit (Logic) reserves the entry for said peripheral.

3. System according to claim 2, in which the logic circuit (Logic) is configured, when a translation requested by a memory request from a peripheral is absent in the buffer memory (TLB), to look for it in the input / output page table (IOPT) by means of a module (PTW) of current page pointer of the tables, and in case of not obtaining the translation in the input / output page table (IOPT), to raise an interrupt to the corresponding core (C).

4. System according to claim 3, in which the logic circuit (Logic) is configured, in case of obtaining the translation in the input / output page table (IOPT), to search for the presence of an invalid entry in the buffer memory (TLB), and in case of presence of an invalid entry in the buffer memory (TLB), to allocate the invalid entry of the buffer memory (TLB) to said translation and write it there.

5. System according to claim 3 or 4, wherein, the logic circuit (Logic) is configured, in case of obtaining the translation in the input / output page table (IOPT), to search for the presence of an invalid entry in the buffer memory (TLB), and in case of absence of an invalid entry in the buffer memory (TLB), to search for the presence of an unreserved entry whose reservation bit (Res) corresponds to a non-reservation in the buffer memory (TLB), and in case of presence of an unreserved entry in the buffer memory (TLB), to allocate the unreserved entry of the buffer memory (TLB) to said translation and write it there.

6. System according to one of claims 3 to 5, in which, the logic circuit (Logic) is configured, in case of obtaining the translation in the input / output page table (IOPT), to search for the presence of an invalid entry in the buffer memory (TLB), and in case of absence of an invalid entry in the buffer memory (TLB), to search for the presence of a non-reserved entry whose reservation bit (Res) corresponds to a non-reservation in the buffer memory (TLB), and in case of absence of a non-reserved entry in the buffer memory (TLB), to search for the presence of a reserved entry whose reservation bit (Res) corresponds to a reservation in the buffer memory (TLB) whose transaction is being processed, to allocate the reserved entry whose transaction is being processed to said translation and write it there.

7. System according to one of the preceding claims, in which the unit (IOMMU) for managing peripheral access (PER) to the memory The shared SoC is configured to be reserved for devices that are not predefined as trusted.

8. System according to one of the preceding claims, in which the processor is also provided with at least one memory management unit (MMU), and a memory management unit page table (MMUPT).